File size: 6,297 Bytes
bd5a354 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 |
import queue
import re
import threading
from concurrent.futures import ThreadPoolExecutor
from typing import Dict, Optional
import cv2
import easyocr
import numpy as np
import torch
import torch.nn as nn
class LicensePlateReader(nn.Module):
def __init__(
self,
model,
char_to_num_mappings: Optional[Dict[str, str]] = None,
num_to_char_mappings: Optional[Dict[str, str]] = None,
confidence: float = 0.30,
queue_size: int = 10,
):
"""
Initialize the LicensePlateReader with the given model and mappings.
Args:.
model OCR model for reading text.
char_to_num_mappings Mappings from characters to numbers.
num_to_char_mappings Mappings from numbers to characters.
confidence threshold for accepting OCR results.
"""
super(LicensePlateReader, self).__init__()
# Initializing
self.char_to_num_mappings = char_to_num_mappings or {
"L": "4",
"D": "0",
"S": "5",
"Z": "2",
"B": "8",
"C": "0",
}
self.num_to_char_mappings = num_to_char_mappings or {
"2": "Z",
"4": "A",
"6": "G",
"5": "S",
"0": "D",
"7": "T",
"8": "B",
}
self.model = model
self.confidence = confidence
self.input_queue = queue.Queue(maxsize=queue_size)
self.output_queue = queue.Queue(maxsize=queue_size)
self.executor = ThreadPoolExecutor(max_workers=2)
self.processing_thread = threading.Thread(
target=self._process_queue, daemon=True
)
self.processing_thread.start()
def forward(self, numbers_side: np.ndarray, letters_side: np.ndarray) -> str:
self.input_queue.put((numbers_side, letters_side))
return self.output_queue.get()
def _process_queue(self):
while True:
numbers_side, letters_side = self.input_queue.get()
result = self._process_single_plate(numbers_side, letters_side)
self.output_queue.put(result)
self.input_queue.task_done()
def _process_single_plate(
self, numbers_side: np.ndarray, letters_side: np.ndarray
) -> str:
future_preprocessed_numbers = self.executor.submit(
self._pre_process, numbers_side
)
future_preprocessed_letters = self.executor.submit(
self._pre_process, letters_side
)
preprocessed_numbers_side = future_preprocessed_numbers.result()
preprocessed_letters_side = future_preprocessed_letters.result()
future_extracted_numbers = self.executor.submit(
self.predict, preprocessed_numbers_side
)
future_extracted_letters = self.executor.submit(
self.predict, preprocessed_letters_side
)
extracted_numbers_side = future_extracted_numbers.result()
extracted_letters_side = future_extracted_letters.result()
future_postprocessed_numbers = self.executor.submit(
self._post_process, extracted_numbers_side, True
)
future_postprocessed_letters = self.executor.submit(
self._post_process, extracted_letters_side, False
)
postprocessed_numbers_side = future_postprocessed_numbers.result()
postprocessed_letters_side = future_postprocessed_letters.result()
return postprocessed_numbers_side + "" + postprocessed_letters_side
def _pre_process(self, frame: np.ndarray) -> np.ndarray:
"""
Preprocess the input frame by blurring, grayscaling, and thresholding.
Args:
frame Input image frame.
Returns:
Preprocessed binary image.
"""
# Blurring
# blurred_frame = cv2.GaussianBlur(frame, (3, 3), 0)
# greyscaleing
greyscaled_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# thresholding white and black
_, binary_frame = cv2.threshold(
greyscaled_frame, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU
)
return binary_frame
def predict(self, frame: np.ndarray) -> str:
"""
Predict text from the preprocessed frame using the OCR model.
Args:
frame Preprocessed image frame.
Returns:
Extracted text if confidence is above threshold, else an empty string.
"""
# OCR model extraction
extraction = self.model.readtext(frame)
# loop all text extractions that above the confidence
for _, text, confidence in extraction:
if confidence > self.confidence:
return extraction[-1][1]
# Error handling
return "" # raise Exception No OCR reading
def _post_process(self, extracted_text: str, is_numbers: bool) -> str:
if not extracted_text:
return ""
if is_numbers:
result = extracted_text.strip()
result = "".join(
self.char_to_num_mappings.get(char, char) for char in result
)
result = "".join(re.findall(r"\b([0-9]{1,4})\b", result))
# if not result.isdigit() or len(result) > 4:
# return ""
return result
else:
result = extracted_text.strip().upper()
result = "".join(
self.num_to_char_mappings.get(char, char) for char in result
)
result = "".join(re.findall(r"[A-Z]{3}", result))
if len(result) != 3:
return ""
return result
def annotate_frame(self, frame, bbox, extracted_text):
if bbox is not None:
color = (0, 255, 0) if extracted_text else (0, 0, 255)
label = "No Extraction" if not extracted_text else extracted_text
cv2.rectangle(frame, (bbox[0], bbox[1]), (bbox[2], bbox[3]), color, 2)
cv2.putText(
frame,
f"{label}",
(bbox[0], bbox[1] - 10), # Top left above the bounding box
cv2.FONT_HERSHEY_SIMPLEX,
0.9,
color,
2,
)
return frame
|